We can offer on site training for your team, to quickly get a foot ahead in the field of deep learning. We can tailor training to your needs and the experience of your team. We can introduce you into the field of artificial intelligence and more specifically deep learning and neural networks; cover the basics of neural network usage and training, including the details of the back-propagation algorithm; look into specific neural network types and their caveats; and get you started with hands-on labs on popular deep learning frameworks, high level libraries, our own byET-engine, data aquisition, data augumentation, neural network training, and finally deployment, from idea until a running system.
consulting – machine learning – neural network synthesis
We can help you design, train, optimize and deploy your deep learning system. We have the necessary expertice and experience.
The first step when developing a deep learning system is to design the neural network. We evaluate what is the appropriate neural network architecture needed for the problem domain and the constrains of the project. Then we adjust the network parameters, to tailor it to your project. We also implement the required additional algorithms for processing the input and output to the deep neural network.
During the training phase we apply our knowledge to gather the optimal training, testing and verification data. We use different data augumentation techniques to improve on the raw data and achieve better results. We run the training of the neural network on one or more servers with hardware acceleration, to reduce the amount of time necessary.
automated machine learning
We use machine learning to produce better deep neural networks. We explore the problem space algorithmically to find optimal solutions beyond of what is attainable by human experts.
We know how to make deep neural networks work efficiently on embedded systems. We understand how to reduce to a minimum the footprint of the neural network, the dependent software and the execution environment. We know how to get the most performance of different embedded CPUs, using the available hardware for speeding up calculations.